Logo image
Back
Centering Survivor Agency in AI for Violence Prevention
Journal article   Peer reviewed

Centering Survivor Agency in AI for Violence Prevention

Aislinn Conrad
Journal of the Society for Social Work and Research
04/29/2026
DOI: 10.1086/742111

View Online

Abstract

Background and Purpose: Artificial intelligence (AI) is increasingly embedded in violence prevention, including child welfare, intimate partner violence, and elder abuse, often without survivor knowledge or consent. Framed as neutral tools, algorithmic systems can reproduce structural inequities when administrative data encode poverty-linked system contact and racialized surveillance. This conceptual paper advances a survivor-centered framework for evaluating ethical AI in violence prevention, defining survivor agency as meaningful voice, transparency, and contestability within institutional decision-making. Methods: Drawing on interdisciplinary scholarship, the paper develops a conceptual framework for survivor-centered ethical evaluation of AI in violence prevention. Results: The framework identifies three mechanisms translating structural disadvantage into algorithmic harm: carceral logic, in which risk classification authorizes expanded monitoring; structural bias, in which inequitable data produces disproportionate risk-labeling of marginalized families; and epistemic injustice, in which lived expertise is discounted relative to opaque algorithmic outputs. From these, the paper proposes three ethical thresholds: Survivor Agency Over Automation, Bias Audits as Ethical Imperatives, and Redistributive AI Conclusions: Rather than endorsing technological adoption, the framework specifies conditions under which algorithmic tools should be constrained, redesigned, or refused, with policy levers including impact assessments, transparency requirements, and survivor participation throughout evaluation.
artificial intelligence social work ethics intimate partner violence child welfare

Details

Metrics

1 Record Views
Logo image